为了能够更快速逼真地模拟室外自然条件下的虚拟场景,提出了一种基于真实拍摄的单目视频的雾天场景模拟算法.算法首先对初始输入图像中场景进行语义分割,针对不同类型的场景进行深度信息的解析;同时对多帧图像序列之间的光流场分布情况和摄像机的运动参数进行估计,完成不同帧图像之间的深度映射与修正,最后基于大气散射模型的原理进行雾天场景的模拟与仿真.理论分析和实验表明,所提出的算法与传统的基于虚拟现实的方法相比效率大大提高,且生成的雾天虚拟场景更具真实感.
A new algorithm based on a real image is proposed to simulate the foggy scene.Firstly,it segments the input image into several regions with semantic,which labels the region as sky,ground and the targets on the ground.Secondly,it gets accurate result from estimation in depth for different scenes.At the same time,the optical flow is computed and information in depth is mapped between the frames in the video.Finally,it gives the simulation of virtual foggy scenes under the atmospheric scattering model.The theory and experiment show that the efficiency of the proposed algorithm is only related to the resolution of the image,and the synthetic foggy scenes are more realistic.